AlphaGo Creator Raises $5.1B to Build Superintelligence
April 27, 2026 — London. Ineffable Intelligence launched today, with Sequoia Capital confirming it co-led a $1.1 billion funding round at a $5.1 billion post-money valuation. Nvidia's venture arm contributed at least $250 million. DST, Index, BOND, the British Business Bank, and Wellcome Trust also participated. David Silver — the British researcher behind AlphaGo, AlphaZero, and AlphaStar at Google DeepMind — is founder and director. The company's website went live this morning with a single stated mission: "Making first contact with superintelligence." There is no product, no revenue, no public roadmap.
The Bet Sequoia Flew to London to Make
In February, Bloomberg reported a roughly $1 billion raise at ~$4 billion pre-money. Today's final close comes in higher: $5.1 billion post-money. Sequoia managing partner Alfred Lin and partner Sonya Huang flew personally to London to secure the deal — unusual theatre from a firm that sees thousands of pitches a year. The company was incorporated in November 2025. Silver was appointed director on 16 January 2026. The UK Companies House filing confirms he holds 75% or more of shares and voting rights.
The speed and scale of the capital, deployed before a single benchmark, demo, or architecture disclosure, is itself the story. This is not a startup priced on revenue multiples. It is a startup priced on one man's résumé and one thesis about the future of intelligence.
What David Silver Actually Built — and Why It Matters
Silver's track record is not incidental context. It is the entire underwriting.
AlphaGo, released in 2016, was the first system to defeat a professional Go player without handicap. Go has roughly 10^170 legal positions — more than the estimated 10^80 atoms in the observable universe — putting brute-force computation permanently out of reach. In March 2016, AlphaGo defeated 18-time world champion Lee Sedol 4–1 in a match watched by 200 million people across Asia. Marc Andreessen called it AI's "Sputnik moment." The investment wave it triggered runs through to DeepMind's Nobel Prize in Chemistry for AlphaFold.
AlphaZero then mastered Go, chess, and shogi simultaneously through pure self-play — zero human game data — raising its Elo from around 3,700 to over 5,000. AlphaStar reached grandmaster level in StarCraft II against professional players. AlphaProof proved three of five problems at the 2024 International Mathematical Olympiad, hitting silver-medal performance in formal mathematics.
The common thread: each system achieved non-human performance by learning from consequences rather than copying human examples. AlphaGo's famous Move 37 in Game 2 against Lee Sedol appeared in no human game record. The machine had reasoned past human intuition.
Human Data Is a Ceiling, Not a Foundation
In April 2025, Silver co-authored "Welcome to the Era of Experience" with Richard Sutton — the Alberta researcher widely regarded as the father of reinforcement learning. Their argument: large language models are structurally capped because they learn exclusively from human-generated data. They can synthesise, extend, and recombine existing knowledge. They cannot discover anything genuinely new.
Ineffable's proposed alternative is what Sequoia calls a "superlearner" — an agent that acquires knowledge without pre-training or imitation, purely through interaction with purpose-built environments. The aspiration is a system that could derive physical law from first principles, invent new branches of mathematics, or design materials beyond current scientific vocabulary.
Silver himself calls human data "fossil fuel": finite, increasingly depleted at the frontier, prejudged by the limits of the cognition that produced it.
The Hard Question Investors Must Sit With
The most important thing to understand about this round is not what Ineffable has built. It is what the round reveals about how frontier AI capital now operates — and where the sharpest risk lies.
Sequoia and Nvidia can rationally justify a $5.1 billion option: missing the post-LLM paradigm is more dangerous to their portfolios than overpaying for one bet on it. For ordinary investors attempting to replicate that logic, the calculus is structurally different. They will not receive the same information rights, pro-rata access, or portfolio-level payoff structure.
The deeper technical problem is one the promotional language quietly skips: what is the environment? AlphaZero succeeded because Go offers complete rules, exact reward, cheap simulation, and no ambiguity about the objective. General intelligence has almost none of those properties. Reward design — not compute, not talent — is the central bottleneck. An agent optimising a proxy reward in an open-ended domain does not discover superintelligence; it games a benchmark. And every environment Ineffable builds will encode human knowledge in its structure, making "no human data" a philosophical position rather than a literal technical constraint.
The valuation is rational only under paradigm-change probability logic. Under normal startup underwriting, it is not. Ineffable is a credible scientific moonshot priced like a de-risked category winner.
The mismatch is the whole story.
not investment advice